Copyright

by

Robert Wayne Ressler

2015

The Thesis Committee for Robert Wayne Ressler

Certifies that this is the approved version of the following thesis:

The Role of Educational Nonprofits in the Early School Achievement of

Children from Diverse Backgrounds

APPROVED BY

SUPERVISING COMMITTEE:

Robert Crosnoe

Pamela Paxton

Supervisor:

The Role of Educational Nonprofits in the Early School Achievement of

Children from Diverse Backgrounds

by

Robert Wayne Ressler, BA

Thesis

Presented to the Faculty of the Graduate School of

The University of Texas at Austin

in Partial Fulfillment

of the Requirements

for the Degree of

Master of Arts

The University of Texas at Austin

May 2015

iv

Acknowledgements

This thesis would not have been possible without the help of Robert Crosnoe,

Pamela Paxton, Chandra Muller, Carmen Gutierrez, the staff and resources of the

Population Research Center and my friends and family.

v

Abstract

The Role of Educational Nonprofits in the Early School Achievement of

Children from Diverse Backgrounds

Robert Wayne Ressler, MA

The University of Texas at Austin, 2015

Supervisor: Robert Crosnoe

Nonprofit organizations represent a potentially powerful source of intervention

for struggling public educational services, yet little is understood concerning the

relationship between nonprofit providers and the educational outcomes of children. This

study uses national data sets from the National Center for Educational Statistics and the

National Center for Charitable Statistics to examine the associations between three

theoretically derived nonprofit measures (competitor, intervener, and youth developer

nonprofits) and student academic outcomes in math and reading over the crucial school-

entry transition period. Results indicate that the number of nonprofits in a community

display some positive associations with math and reading score gains, but that these

associations must be carefully interpreted in regards to the heterogeneity of nonprofit

service and the socio-economic context of the child.

vi

Table of Contents

List of Tables…………………………………………………………………vii

List of Figures………………………………………………………………...viii

TEXT.………………………………………………………………………..... 1

Appendix A Fully Controlled Models Predicting Math and Reading Score

Gains…………………………………………………………………..27

References……………………………………………………………………..39

Vita…………………………………………………………………………….45

vii

List of Tables

Table 1: Descriptive Statistics for Prekindergarten and Kindergarten Data…...29

Table 2: Mean IRT Scores by Number of Nonprofits by Groups……………..31

Table 3: Results of Models Predicting Math Scores by Nonprofit Counts……32

Table 4: Results of Models Predicting Reading Scores by Nonprofit Counts…33

viii

List of Figures

Figure 1: Influences on Child Outcomes during the Transition to School..34

Figure 2: Predicted Kindergarten Math Test Scores in ECLS-K,

by Competitor Nonprofits and Family Income…………………….35

Figure 3: Predicted Kindergarten Math Test Scores in ECLS-K,

by Youth Developer Nonprofits and Family Income………………36

Figure 4: Predicted School Entry Reading Test Scores in ECLS-B,

by Youth Developer Nonprofits and Family Income……………….37

Figure 5: Predicted Kindergarten Reading Test Scores in ECLS-K,

by Intervener Nonprofits and Family Income………………………38

1

The challenges that American schools face today are wide ranging, from

underperformance in science and math (National Science Board 2014) to the

disproportionate attainment of students from diverse backgrounds (Duncan and Murnane

2014). Innovative solutions are desperately needed to meet these challenges. One area

that shows particular promise concerns intervention during the years surrounding a

child’s transition into school, which are the foundation of the entire educational career

and a driving force in educational disparities. As such, investments in students during this

period return higher rewards than at other points in the educational timeline (Heckman,

2006; Alexander, Entwisle, and Olson 2007). The best way to provide these investments,

however, is often less understood.

One understudied source of investments in this growing field of research concerns

the potential for outside organizations, such as nonprofits, to support the educational

mission of schools. This lack of attention to the nonprofit sector is notable for theoretical

reasons, given the emphasis of general systems theory on the interconnected nature of

institutional and ecological actors in individual outcomes (Bertalanffy 1969). It is also

notable for practical reasons, given the increased attention to school-community

partnerships as a means of addressing educational problems in major educational policies

such as No Child Left Behind (NCLB; U.S. Department of Education 2002) and the

growing awareness that the effectiveness of government and charitable spending needs to

be better assessed (Garrett and Rhine 2010).

In this spirit, this project integrates underutilized governmental data with widely

used educational data to examine the potential for the presence of youth-focused

2

educational nonprofits in a community to promote student success in that community’s

schools. Specifically, it adds zip code-level data on the quantity and type of educational

nonprofits to the Early Childhood Longitudinal Study-Birth Cohort (ECLS-B) and

Kindergarten Cohort (ECLS-K), applies multilevel modeling to these combined data sets,

and examines whether the nonprofit makeup of a community predicts growth in

achievement and reduces socioeconomic disparities in such growth during and after the

start of elementary school. Doing so answers the challenge from Pianta and Walsh (1996,

p. 76) for researchers to develop “a coherent set of principles…to understand the activity

of [the systems of education], how they behave, how they change, and ultimately, how to

predict their functioning with respect to the outcomes of schooling.”

The larger context of this study is the flow of funding (measuring billions of

dollars) from government, businesses, and individuals to nonprofits aiming to reduce

social problems (Roeger, Blackwood, and Pettijohn 2012; Epstein and Buhovac 2009).

When individuals and business utilize tax deductible donations, this directly removes

money from public coffers and often does not redistribute it efficiently (Reich 2005). In

general, this revenue flow diverts funding from improving public services, often with the

implicit or explicit argument that due to market characteristics these organizations are

more effective than those public services (Anheier 2005; Clotfelter 1992; Odendahl

1991). Importantly, this argument has been subjected to little systematic investigation

concerning the effectiveness of nonprofits and other similar organizations, especially

during the critical school transition period. This study, therefore, takes some of the first

steps necessary to formulate a better understanding of these often forgotten

3

organizational factors that may influence children’s success in school at a time when it

has such far-reaching consequences for them and society at large.

The Educational Problem

Recently, the U.S. educational system has been criticized as falling behind many

other developed countries in producing highly skilled students for the global workforce

(National Science Board 2014). It also appears to be complicit in the reproduction of

inequality, as evidenced by the enduring racial/ethnic and socioeconomic disparities in

numerous educational outcomes that public schooling is supposed to reduce

(Darensbourg and Blake 2014, Bates and Glick 2013, Education at a Glace 2007,

Ackerman, Brown, and Izard 2004, Jencks et al. 1972). Importantly, these disparities

exist before children even enter school. Children from low-income families, for example,

score significantly lower than children from middle- and upper-class homes on math and

reading tests at the beginning of kindergarten (Lee and Burkam 2002). These initial

disparities then increase over time as children move through the educational system and

are subjected to stratifying forces like ability grouping, teacher expectations, and

between-school differences in quality (Alexander et al. 2007). Thus, the window

surrounding the transition into school—the year before the start of kindergarten and the

kindergarten year itself—makes up one of the most important periods in students’ long-

term educational trajectories and represents a critical point for interventions aiming to

improve the academic prospects of children and reduce disparities among child groups

(Pianta, Cox, and Snow 2007, Varnhagen, Morrison, and Everall 1994). If the public

educational system is going to address this reproduction of inequality and boost its

4

overall effectiveness, the years surrounding the transition into formal schooling are of the

upmost importance.

These pressing issues confronting the public educational system necessitate

innovative solutions that will require input from numerous actors, both inside and outside

of schools. This need for school-community partnerships was recognized twenty years

ago when the 1994 Congress passed the Goals 2000 report (Epstein 1995) and then

reiterated in the NCLB legislation (NCLB 2002) as well as in numerous state-level

educational practices (National Education Association 2011). This policy

conceptualization of communities as partners with the institutions serving them dovetails

with sociological conceptualizations of communities as elastic political constructs that

involve the active participation of individuals in order to maintain real existence (Collins

2010). I argue that nonprofit organizations—defined by the federal government and the

National Center for Charitable Statistics as tax exempt 501c3 organizations (NCCS

2002)—embody both conceptualizations of community. First, they can be partners

interested in investing in and supporting an institutional structure that serves the well-

being of children in a community. Second, they can be active participants in the social

defining of the communities that they both draw on and serve.

Leaving aside the unfortunate dearth of theoretical analysis concerning the

specific role of such third-party actors within the educational system, nonprofits represent

a potentially powerful vehicle for policy intervention in this system, especially in an era

of shrinking government programs supporting children and their educational endeavors

(Clotfelter 1992). This logic is reflected in economic perspectives that position

5

nonprofits, along with the private and public sectors, as a channel of resource

mobilization through which social problems can be addressed (Weisbrod 1972). Because

the nonprofit sector is supported by both government and business and has been

historically portrayed as a counterweight to failing public services (Clotfelter 1992), it is

a meeting point in this synergistic resource mobilization within communities.

Additionally, scholars have linked nonprofits to community socioeconomic

characteristics (Katz 2014; Allard 2009) political culture (McDougle and Lam 2014,

Bielefeld 2000) and altruism (Rose-Ackerman 1996) with an increased focus on the

importance of nonprofit service provision in today’s modern economy (Anheier 2014;

Salamon, Hems, and Chinnock 2000). In the educational context, nonprofits that focus on

improving the outcomes of school-aged children may represent a critical component for

the success of the public education system, or, more troublingly, their presence may only

exacerbate existing inequalities (Odendahl 1991). This research represents a first pass at

attempting to evaluate this public-nonprofit relationship on the national level.

Aims of the Study

Utilizing a conceptual model in which nonprofits represent a significant

proportion of “community partners” that interact with public schools and families to

promote child-wellbeing (see Figure 1), this study aims to examine the diverse ways that

nonprofits in a community may support the educational mission of schools in that

community. The general goal is to examine the association between the community

presence of educational nonprofits and the test scores of the children as they transition

6

into elementary school. More specific goals then concern potential policy relevant

sources of variability in this basic association.

[Figure 1 About Here]

To begin with the general goal, the nonprofit makeup of a community might be

associated with the academic performance of students in a community’s schools through

two separate but not mutually exclusive mechanisms: competition and supplementary

support. Beginning with competition, educational organizations and activities that offer

alternatives to traditional public schools may trigger general educational improvements

across the board by stimulating innovation through competition (Zimmer and Buddin

2009). In other words, if nonprofits provide alternative educational opportunities to

schools, they can spur improvements in both private and public options that eventually

lead to better performance across the board. For example, this logic has been applied to

investigations of the impact of charter schools on traditional public schools and while

there is some contention regarding the true nature of this impact some studies have

revealed positive effects of charter schools on public school outcomes (Sass 2006;

Bettinger 2005). Turning to supplementary support, many nonprofits’ missions are to

promote the success of community members, thereby directly and indirectly adding onto

(or complementing) school services for children. For example, Leventhal, Dupéré, and

Shuey (forthcoming: 152) argue that, “the quantity, quality, diversity, and affordability of

programs and resources at the neighborhood level are an aspect of neighborhoods that is

likely to be important for child development, as well as a potential pathway through

which neighborhood structural characteristics may influence child development.” If there

7

are more nonprofits in an area specifically organized to improve child educational

outcomes, then some impact of these neighborhood resources on children’s actual

outcomes would be expected.

This study uses the National Taxonomy of Exempt Entities (NTEE) to tease apart

these potential explanations for any observed association between community-based

nonprofits and the outcomes of public school students by grouping nonprofit

organizations based on whether they represent separate competitive alternatives to public

schools or provide services to schools. Although their services might differ, the roles of

both competing and supplementary support organizations in children’s early achievement

are theoretically similar, leading to Hypothesis 1: The greater the number of nonprofit

organizations providing services, whether in competition with or in support of the public

education system, the more that student performance will improve.

Turning to issues of variability, one possibility is that nonprofit interactions with

public schools could be more successful (i.e., associated with higher achievement) in

some subject areas over others. Previous research from many fields has found substantial

differences between the impacts of a host of independent variables (e.g., cash incentives,

NCLB, social capital, etc.) and student achievement across academic subjects like math,

reading, and science (Bettinger 2012; Dee and Jacob 2010; Leana and Pil 2006; McKown

and Weinstein 2002; Stevenson, Schiller, and Schneider 1994). Reading is a subject of

intense focus in the early childhood education years and in the primary grades of

elementary school (Sénéchal and Young, 2008; Armbruster, Lehr, and Osborn 2001;

Snow, Burns, and Griffin 1998; Mason 1980). Any nonprofit community partner working

8

with schools should also be attuned to the importance placed on reading and therefore

may be more focused on improving children’s language and literacy skills rather than

their development of math skills. The documented success of one-on-one reading

interventions may also contribute to nonprofits focusing on reading interventions

specifically (Ritter, Barnett, Denny and Albin 2009; Invernizzi, Rosemary, Juel and

Richards, 1997). These potential motivating factors lead to Hypothesis 2: The presence of

nonprofits will be more strongly associated with children’s reading achievement than

with their math achievement.

Another potential source of variability concerns timing within the transition into

elementary school. The first component of this transition is how school ready children

are; in other words, what level of academic skills they bring into formal schooling at the

start of the kindergarten year and their early childhood trajectory of learning up to that

point. The second component concerns what happens to children once school starts; in

other words, what skills are gained in relation to their initial level of school readiness

across the kindergarten year, indicating what their future learning trajectories are likely to

be (Crosnoe, Bonazzo, and Wu 2015). The skill begets skills perspective posits a highly

cumulative process of learning that prioritizes the value of early intervention (see

Heckman 2000). This argument, and supporting evidence of the potentially high returns

to early childhood interventions, suggests the value of taking action to promote the

human capital development of children before the start of school. One reason is that who

attends early childhood education programs (which are voluntary and often expensive)

varies far more widely than who attends school (which is mandatory and free), so that

9

interventions targeting these early years have potential to even out basic issues of access

and opportunity (Duncan and Magnuson 2013; Shonkoff and Phillips 2000). Thus, I pose

Hypothesis 3: The presence of nonprofits will be more strongly associated with academic

achievement around kindergarten entry than during the kindergarten year.

A final source of variability concerns the difference between overall performance

levels and sociodemographic disparities in performance. Major educational policies

typically target both—improving how much all students learn while also reducing

disparities in rates of learning across groups. To do both, a policy would need to facilitate

skill development across the board but more so for traditionally disadvantaged groups

than more advantaged groups. Although not always borne out in reality, the theoretical

argument for this two-pronged philosophy is that the infusion of resources and supports

will matter more to students who have few resources overall or would make more of a

difference to them than they would for fellow students for whom multiple resources are

redundant. Thus, the child experiencing disadvantage and the child with advantages both

move ahead, with the former closing some of the distance on the latter in the process

(Crosnoe and Benner 2015; Ceci and Papierno 2005; Arum 2000). This dual philosophy

is central to the policies surrounding K-12 schooling, such as No Child Left Behind,

while reducing disparities has been more in the spotlight in the recent push to support the

expansion of early childhood education (Fuller 2007; Barnett and Belfield 2006). Given

the importance of reducing disparities in the transition into school years, determining if

nonprofits coincide with a reduction in disparities across groups is a crucial goal.

Following this logic, I propose Hypothesis 4: The presence of nonprofit organizations

10

will be most strongly associated with academic achievement of children from low-income

families.

Testing these hypotheses within the general systems conceptual model in Figure 1

is important because of the extant lack of consideration of the impact of nonprofit

organizations on a national level. Incorporating nonprofits into the theoretical models of

community systems in which families, children, and schools are located could provide

valuable insight into potential mechanisms to improve student achievement and reduce

systemic disparities, helping to translate sociological research into policy action.

Methods

Data

The two sources of nationally representative child-level data used in this study

were both collected by the National Center for Education Statistics (NCES). First, ECLS-

B (see Snow et al. 2007) is a nationally representative sample of 10,700 children born in

the U.S. in 2001 who were followed from nine months through kindergarten entry (2006

or 2007). Data were collected in multiple ways, including interviews with parents,

caregivers, and teachers and direct assessments of children. The analytical sample used

here included all children who participated in the age 4 and kindergarten waves who had

direct assessments and zip code information (n = 6,320; note, per NCES reporting

requirements, all sample sizes are rounded to the nearest 10). Second, ECLS-K is a

nationally representative cohort of over 21,000 children enrolled in approximately 1,000

schools in 1998 (Rathbun and West 2004). The multistage sampling frame began with

100 primary sampling units comprising counties and county groups from which schools

11

were sampled, with approximately 23 students from each school selected (West, Denton,

and Reaney 2000). ECLS-K also includes data from interviews with parents and direct

assessments of children. The analytical sample used here included all children who

participated in both kindergarten waves who had zip code information (n = 16,460).

For both data sets, I included all available data on children whether they attended

public or private school in kindergarten. This inclusion maintained the representativeness

of the sample, maximized sample size, and allowed the nonprofit results to be

generalizable to all U.S. children. Given my conceptual focus on public education,

however, I controlled for school sector and also did a comparative analyses for the public

school subsample (n = 5,400 ECLS-B; n = 12,840 ECLS-K), which, not surprisingly,

revealed that this large subsample drove the results reported for the full sample when

controlling for school sector.

As explained below, longitudinal sampling weights accounted for differential

attrition across waves in both data sets, and missing data estimation retained all cases

within the two analytical samples. Finally, the zip code-level data to be merged into these

two child-level data sets came from the Master File of nonprofit data from the NCCS, a

cumulative list of all exempt organizations from 1989 to present and consists of over 2.5

million nonprofits, their addresses, years nonprofit status was received, and latest tax

filing years.

Measures

Children’s achievement. ECLS-B and ECLS-K assessed children in reading and

math using individually administered two-stage adaptive tests, with content areas and

12

domains based on the National Assessment of Educational Progress framework (NCES

2001). Used here are Item Response Theory (IRT) scale scores, which estimate patterns

of responses for questions based on patterns of right, wrong, and omitted responses and

on item parameters of difficulty, discriminating ability, and “guess-ability” (Rock and

Pollack 2002). The two surveys were designed to have comparable measures. More

information about the measurements can be found in the codebook (NCES, 2001) and a

report from Rock and Pollack (2002). In ECLS-B, I used the test scores from

kindergarten entry point as the dependent variable and the test scores from the prior pre-

kindergarten wave as a lagged independent variable. In ECLS-K, I used the test scores

from the spring of kindergarten and the fall of kindergarten in the same way. Descriptive

statistics for these test scores (and all other variables) are presented in Table 1.

Community nonprofit composition. Three variables captured the total number of

nonprofits (categorized by their NTEE codes) that were registered in a community area,

defined through mapping software as a zip code and its contiguous neighbors. These

community areas were also used to construct some other variables listed below.

Competitors consisted of registered nonprofit preschools, primary, and elementary

schools as well as elementary charter schools in a child’s community area. The second

two variables are subsets of the supporter category. The variable for interveners counts all

nonprofit educational services, remedial reading, and encouragement organizations as

well as all parent teacher groups in the community area. The variable for youth

developers counts supportive nonprofit organizations listed under NTEE code “O” for

Youth Development such as youth centers and clubs, adult and child matching programs,

13

and scouting organizations. All counts include only nonprofits that began operation

before the survey data was collected. Using the NTEE codes in this way should also

accomplish the goal of focusing on nonprofit organizations that likely serve the

communities in which they are situated.

Community socioeconomic status. Because of the strong link between the

socioeconomic composition of communities and children’s academic indicators as well as

theoretical links between community SES and nonprofit concentration, I assessed the

socioeconomic status of each zip code with a composite of three standardized items (see

Leventhal and Brooks-Gunn 2000): the percent of individuals in the area with a master’s

degree and above, the percent classified as professionals, and a dichotomous indicator

signaling if the average household income for the zip code was twice the median

household income of the entire sample (α = 0.76 in ECLS-B and 0.80 in ECLS-K).

Other covariates. A number of factors were taken into account to deal with

various confounds related to selection into neighborhoods, nonprofit concentration, and

children’s achievement. Community covariates include urbanicity (1 = urban), region (1

= south), total number of children 6 and under in the community area, and the total

number of square miles of that area. Family covariates include parent education (1 =

highest level of parent education in household is less than high school and 5 = beyond a

bachelor's degree), and low income (1 = total household income at 185% of the federal

poverty line for household size or lower). Child covariates included center or preschool

enrollment in the year before kindergarten (1 = enrolled), age in months, gender (1 =

female), and race (dichotomous indicators for African American, Hispanic, and Other).

14

School and assessment covariates include sector (1 = public), timing of assessment (in

days from first assessment date), and language of assessment (1 = Spanish). Additionally,

in models for the kindergarten year, school covariates include Title 1 receipt (1 = yes),

percent minority students (0-100), and school size (1 = 0-149, to 5 = 750 or more).

Plan of Analysis

The four hypotheses were tested in a series of regression models predicting test

scores at time t by test scores at time t-1 (creating a lagged modeling structure effectively

reducing endogeneity and capturing gains in scores over time; see Glazerman, Levy, and

Myers 2003) and the focal community-level nonprofit predictors, with covariates and

interactions of the nonprofit variables with low-income status added iteratively. The

equation for these models is:

Yi = β0 + β1X1i + β2 X2i + β3X3i + β4X4i + β5X5 +… + βkXk + ei

Models were estimated for each subject area and, within these subjects, for the

period encompassing kindergarten entry (pre-kindergarten wave to the start of

kindergarten in ECLS-B) and for the period encompassing the full kindergarten year (fall

of kindergarten wave to the spring of kindergarten wave in ECLS-K). For each child,

therefore, Yi successively represented the kindergarten math and reading score (ECLS-B)

and the spring kindergarten math and reading score (ECLS-K). β1 represented the effect

of the logged number of competitor nonprofits in a child’s community area on the test

score of interest (X1i), β2 represented the effect of the number of interveners (X2i), and β3

represented the effect of the number of youth developers (X3i). β4 is the effect of the

previous test score (X4i)—prekindergarten score in ECLS-B, fall kindergarten math score

15

in ECLS-K. Finally, β5X5…βkXk represented all additional covariates, with β0 as the

intercept. Due to the potential diminishing returns of the number of nonprofits for

children’s learning (i.e., moving from 0 nonprofits to 1 nonprofit is likely more

meaningful than moving from 20 to 21), all nonprofit variables were logged (after 0.5

was added to each variable to avoid taking the log of zero). Logging changed the

interpretation of the coefficients of interest, so that a one percent increase in the total

number of nonprofits would correspond to an increase or decrease in the gain in a child's

test scores of β/100.

In order to preserve as much data as possible, missing data were accounted for

through mvn multiple imputation in STATA. Given the nested nature of the data (with

children nested within schools within communities), I estimated multilevel models that

explicitly partitioned variances into within- and between-level components. Doing so did

returned no statistically significant improvement of model fit and yielded no

substantively different results than a simpler method of using STATA’s survey

commands to account for the nesting of the data and produce robust standard errors.

Given this lack of difference, the results presented here come from the simpler and more

straightforward STATA approach. Finally, all models include longitudinal sampling

weights to account for cross-wave attrition and to correct for other study design effects

(e.g., the unequal probability of selection into the sample).

Results

As a starting point, Table 2 presents the mean math and reading test scores for

children in each subject in each time period according to the number of nonprofits in their

16

communities. To ease interpretation, the number of nonprofits in a community has been

trichotomized (0, 1-5, 5+) for these descriptive statistics, with significant differences

between the latter two categories and the first category (calculated with t statistics) noted

in the table. During the period encompassing kindergarten entry in ECSL-B, the mean

math test score and the mean reading test score increased as the number of nonprofits (all

categories) increased. During the kindergarten year in ECLS-K, only the mean reading

test score consistently increased in tandem with the number of non-profits (and note that

the math score actually slightly decreased as the number of youth developer nonprofits

increased). Thus, descriptively, the presence of nonprofits in a community did seem to be

associated with children’s achievement in that community, with more consistency up to

school entry rather than after school entry and more consistency in reading than math.

Community Nonprofits and Children’s Math Test Scores

Turning to the hypothesis testing, Table 3 presents the results of a series of

regressions for math test scores by time period. For each time period, Model 1 included

the focal nonprofit factors and the t-1 test score (to create the lagged structure gauging

test score gains), with the full set of covariates added in Model 2, and interactions

(nonprofit variables x family income status) in subsequent models. The Model 2 results

indicate that one nonprofit factor—number of interveners—was associated with

children’s test scores (b = .47, p < .05) after all covariates were controlled. This

significant coefficient indicated that students from communities with three intervener

nonprofits would experience a test score gain 1.4 points greater than those in

communities with no such nonprofits, an effect size equivalent to about 14% of a

17

standard deviation in the school entry math test score distribution in the ECLS-B. This

represents a relatively small effect size, especially compared to the effect sizes of the

other covariates (see Appendix A). For the school entry period, none of the interactions

of the nonprofit factors with family income status were significant at conventional (p <

.05) levels for math (interaction results not included in the table).

The second panel of Table 3 includes the results for the period between the start

and end of kindergarten. The fully controlled results in Model 2 indicated that no

nonprofit factor significantly predicted math test score gains after school had begun, net

of all covariates. Yet, the significant interaction between two nonprofit factors and family

income status (Models 3 and 4) indicated that nonprofits did matter for this subject during

this time period; they just mattered differently for specific segments of the population.

While separate models for interactions between each nonprofit variable and family

income were run, only models with significant interactions are displayed in Table 3 and

Table 4. Models containing all interactions together returned similar results with slightly

reduced statistical significance for both interactions and an additional statistically

significant interaction for youth developer nonprofits for the school entry period. I

present the individually additive models here because they more succinctly represent

overall trends.

To interpret these significant interactions between the number of competitor

nonprofits and family income (b = .43, p < .001) and between the number of youth

developer nonprofits and family income (b = .30, p < .05), I graphed the predicted math

test scores at the end of kindergarten for children who lived in communities with 0

18

nonprofits and those who lived in communities with three nonprofits (roughly one

standard deviation above the mean for those counts), with all other variables in the

model, including the prior test score, held to their sample means. These predicted scores

are presented in Figures 2 and 3.

For both types of nonprofits, children from low-income families posted greater

gains on the math test over time when living in communities with more nonprofits, and

disparities in test scores between such children and their peers from more affluent

families were smaller in those same communities. Technically, these two patterns both

supported Hypothesis 4, which was that children from low-income families would benefit

more from the presence of nonprofits. Yet, this seemingly similar pattern across two

kinds of nonprofits subsumed important differences. The interaction between youth

developer nonprofits and family income status was most clearly in the spirit of the

hypothesis, as the closing of disparities between the two groups of children was due to

the disadvantaged children gaining ground while they more advantaged children did not

lose ground. As Figure 3 shows, children from low-income families gained nearly a point

in communities with three youth developer nonprofits relative to similar children in

communities with none. For children in more affluent families, however, the number of

youth developer nonprofits was not associated with math test score gains.

The interaction between competitor nonprofits and family income status was not

as closely aligned with the spirit of the hypothesis, as the narrowing of disparities was

driven not so much by the gains of children from low-income families as by the losses of

other children. As Figure 3 shows, children from families that were not low-income had

19

lower test scores when living in communities with three competitor nonprofits than in

communities with none (with an opposite, albeit less pronounced, pattern for the children

from low-income families). As a result, test score disparities related to family income

were actually reversed in communities with a greater presence of competitor nonprofits,

but this reversal occurred because of gains by children from low-income families and

losses by children who were not from low-income families.

Because these results were only for math, they do not speak to Hypothesis 2. In

terms of Hypotheses 1, 3, and 4, they contain several relevant observations. First, they

indicate some support for Hypothesis 1, notably that the number of intervener nonprofits

was associated with slight increases in math scores, while the youth developer nonprofits

were associated with increases for children from low-income families. Furthermore,

specifying the type of nonprofit organization proved important for evaluating Hypothesis

1 because intervener organizations showed a positive association with student outcomes

in general while youth developer and competitor nonprofits did not for the sample as a

whole (with competitor nonprofits actually displaying a negative association with math

gains for children not from low-income families). Hypothesis 3 was also partially

supported by the fact that the only significant nonprofit association in the expected

direction in models containing all controls occurred during the kindergarten entry period.

Finally, I found qualified support for Hypothesis 4 in that children from low income

families experienced stronger gains in test scores as the number of competitor and youth

developer nonprofits in their communities increased compared to their peers not from

20

low-income families. For competitor nonprofits, however, this positive trend for children

from low-income families was coupled with a more negative trend for all other children.

Community Nonprofits and Children’s Reading Test Scores

Table 4 presents the results of the same series of regressions for reading test

scores. The fully controlled results from Model 2 in the kindergarten entry period

revealed that no nonprofit factors were significantly associated with reading test scores.

Model 2 for the kindergarten year period did display a significant association between the

number of youth developer nonprofits and reading test scores (b = .34; p <.05). This

significant coefficient indicated that the number of youth developer nonprofits was

positively associated with reading test score gains in the kindergarten year, with another

relatively small effect size. Children in communities with three such nonprofits as

opposed to zero experienced a reading score gain of 1.03 points, an effect equivalent to

about 10% of a standard deviation in the spring kindergarten math test score distribution

in ECLS-K.

Model 3 for both the kindergarten entry and kindergarten year periods revealed

two significant interactions of the nonprofit factors with family income status for this

subject. For the kindergarten entry period, this interaction was between the number of

youth developer nonprofits and family income (b = -0.96, p < .05) and, for the

kindergarten year, it was between the number of intervener nonprofits and family income

(b = -.31, p < .05). Again, to interpret these interaction terms, I graphed the predicted

reading scores for children who lived in communities with 0 and three nonprofits, with all

21

other variables, including the prior test scores, held to their sample means. These

predicted scores are presented in Figures 4 and 5.

Although youth developer nonprofits were associated with gains overall for

children in the kindergarten year period, children from low-income families in the ECLS-

B did not experience gains in reading test scores across the transition into kindergarten

when they lived in communities with greater numbers of youth developer nonprofits.

Instead, their test score gains were weaker in communities with three such nonprofits

than in communities with none, with the opposite (and slightly weaker) pattern for their

peers from families that were not low-income. They did not appear to be more affected

by the presence of youth developer nonprofits than other children, and they certainly did

not appear to be more positively affected by this potential community resource. As a

result, income-related disparities in children’s reading test scores grew in tandem with the

increase in this kind of nonprofit in the community. A similar pattern is also reflected in

the interaction between family income and intervener nonprofits in ECLS-K, with

children from low-income families experiencing no gains in scores from increased

intervener nonprofits and children who were not from low-income families experiencing

about a half a point increase in such communities.

Going back to the hypotheses, in addition to intervener nonprofits being

associated with increased math test scores in the transition into kindergarten, they were

also associated with increased reading test scores in the kindergarten year, (evidence for

Hypothesis 1, mixed support for Hypotheses 2 and 3). No other nonprofit count was

generally associated with reading or math test scores in the child population at large

22

(evidence against Hypothesis 1). Finally, competitor and youth developer nonprofits were

more closely associated with the kindergarten year math test scores of children from low-

income families. Youth developer nonprofits at kindergarten entry and intervener

nonprofits in the kindergarten year, however, were associated with increased reading

score gains from children from families that were not low-income and actually associated

with reduced reading score gains for children in low-income families (mixed support for

Hypotheses 2 and 4).

Conclusion

Nonprofit organizations are a potential mechanism through which interventions in

education can be delivered, despite the fact that they are frequently left out of theoretical

models concerning structural and contextual influences on student outcomes and

educational disparities. The general goal of this research, therefore, was to investigate

whether and how the composition of nonprofits in a community are associated with the

outcomes of the community’s children, a preliminary but necessary step in the

consideration of the usefulness of school-community partnerships that have received so

much public attention.

Consequently, I combined two sources of national data to test several hypotheses

regarding nonprofit involvement in the educational system. I summarize those results

here. Nonprofits appeared to play a diverse role in the achievement gains of some, but not

all, students. For example, the community composition of nonprofit organizations was

more consistently associated with gains on math tests for children from low-income

families, but it was more consistently associated with reading gains for children from

23

families that were not low-income. In terms of timing, links between nonprofit

composition and children’s test scores did not differ overall between the school entry

period and the kindergarten year, but links between nonprofit composition and

socioeconomic disparities in children’s test scores did. Specifically, the presence of

nonprofit organizations seemed to be associated with more optimal patterns of reductions

in disparities (i.e., driven by gains among children from low-income families rather than

by losses among other children) in in the kindergarten year period more than in the

school entry period.

These results underscore three important themes. The first is the importance of

investigating nonprofits according to the theoretically derived functions they may serve

in a community. The differential test score gains experienced by children from low-

income families and other children according to the presence of competitor and youth

developer nonprofits serve to underscore the importance of understanding the

heterogeneity of the nonprofit field – the very heterogeneity that has often been cited as a

major challenge for research on the impact of such organizations. As new sources of data

become available, understanding and exploiting this heterogeneity needs to be an explicit

goal of research. In this study, for example, the presence of youth developer

organizations (nonprofits with missions that would align more with the development of

all youth in a community) was associated with less socioeconomic divergence in math

test score gains while the presence of competitor nonprofit organizations (entities whose

missions would necessarily align more with promoting the success of individuals actually

enrolled in their programs) was not.

24

This point does highlight a limitation of this study: the fact that I cannot

determine if individuals in the NCES data actually participated in nonprofit programs or

not—data that I argue should be gathered in future surveys either at the student or school

level. Similar to the motivation for including children from private schools in the sample,

this data limitation does not interfere with my ability to consider the observable impact of

nonprofit composition on the community as a whole. Furthermore, assessing the

theoretical ability of nonprofit organizations to improve the outcomes of all individuals in

a community—not just direct participants or clients—is one of the motivating factors

behind calls for third sector investment beyond private businesses and the public services

that are perceived to be failing. The very nature of nonprofits serving different

communities at different levels of intervention and intensity necessitates more careful

data collection directly from nonprofits and those they serve rather than indirectly from

tax forms. For now, however, use of the NTEE codes to pair nonprofit organizations

closer to their theoretical missions and the communities they serve represents a

substantial move in this direction.

The second theme concerns the need for a deeper understanding of who is best

served by which nonprofit organizations. This study revealed different associations

between nonprofit measures and the math and reading gains of children from families of

differing levels of income across all time periods. This variability is a challenge, but it

also speaks to the reality of educational policy that one size rarely fits all. More targeted

approaches are likely to be more effective. For example, if intervener and youth

developer nonprofit organizations focus their attention more towards increasing the test

25

score gains of children from low-income families, then they may see more of an impact

from their services. A limitation of this study relevant to this conclusion is that it cannot

explicitly speak to how to target this population or the precise ways to increase their

achievement through nonprofit involvement. What it can do is point future studies to this

possibility to better understand the details and mechanisms behind nonprofit composition

and educational success.

The third theme speaks to issues concerning the association between nonprofit

organizations and the socioeconomic statuses of the individuals who participate in that

their programs. Two unexpected results indicate the importance of theorizing the

different ways individuals from different family backgrounds may interact with nonprofit

organizations. First, the presence of more competitor nonprofits was associated with

increased school entry math scores for children from low-income families but a decrease

in math scores for children not from low-income families during the same period.

Second, when communities housed more intervener and youth developer nonprofit

organizations, socioeconomic disparities in reading scores were wider. Perhaps these

patterns reflect socioeconomic differences in how children and families—and the schools

serving them—interact with nonprofits. Currently, the data are not there to examine these

interactions, suggesting the need for qualitative data collection. Another possibility is

selection—nonprofits clustering in areas that already have problems with disparities.

New kinds of statistical approaches, such as instrumental variables, are needed to delve

more deeply into this possibility.

26

Given the exponential increase in nonprofit organizations over the past twenty

years, these types of non-governmental agencies will be involved in the lives of many

people, especially children, in the years to come. If the sheer number of organizations or

the quantity of money spent on them is not motivation enough, the logic of market

competition suggests that they may be some of the most innovative organizations

focusing their attention on alleviating educational problems and social ills more broadly.

Building on this preliminary work to better situate nonprofits in theoretical frameworks

of educational inequality and collecting the data to directly examine such frameworks,

therefore, should be a goal of sociological research moving forward.

27

Appendix A

Table A1. Fully Controlled Models Predicting Math and Reading Score Gains

School Entry (ECLS-B)

Kindergarten Year (ECLS-K)

Math Reading

Math Reading

Nonprofit Count

Competitors -0.240 -0.616+

-0.054 -0.055

(0.257) (0.368)

(0.088) (0.137)

Supporters

Interveners 0.465* 0.466

-0.066 -0.011

(0.211) (0.311)

(0.084) (0.132)

Youth developers -0.059 -0.051

0.118 0.343*

(0.225) (0.332)

(0.095) (0.145)

Previous Score 0.212*** 0.181***

0.923*** 0.563***

(0.019) (0.024)

(0.011) (0.018)

Low Income (1 = yes) -2.043*** -2.798***

-0.313* -1.793***

(0.383) (0.583)

(0.122) (0.164)

Community SES 1.056*** 1.017*

-0.164 0.072

(0.254) (0.398)

(0.102) (0.162)

Community Covariates

Urbanicity (1 = urban) -0.192 0.798

0.134 0.005

(0.537) (0.728)

(0.171) (0.266)

Region (1 = south) 0.020 2.030**

0.656** 1.256***

(0.377) (0.624)

(0.234) (0.336)

Total children -0.000 0.000

0.000+ -0.000

(0.000) (0.000)

(0.000) (0.000)

Total sqmi -0.000 -0.000

0.000 0.000**

(0.000) (0.000)

(0.000) (0.000)

Race

African American -2.053*** -0.661

-1.419*** -0.468

(0.471) (0.651)

(0.228) (0.297)

Hispanic -2.586*** -2.424***

-0.260 -2.686***

(0.458) (0.625)

(0.206) (0.358)

Other -1.258 -1.150

1.051*** 0.556

(0.779) (1.002)

(0.308) (0.338)

Standard errors in parentheses

*** p<0.001, ** p<0.01, * p<0.05, + p<0.1

Table continued on next page.

28

Table A1 (cont). Fully Controlled Models Predicting Math and Reading Score Gains

School Entry (ECLS-B) Kindergarten Year (ECLS-K)

Math Reading Math Reading

Child Covariates

Parental education 0.953*** 1.699***

0.347*** 1.209***

(0.097) (0.140)

(0.058) (0.082)

Gender (1 = female) 0.293 1.509***

-0.044 1.260***

(0.260) (0.411)

(0.091) (0.135)

Age 0.607*** 0.764***

0.033* 0.151***

(0.036) (0.048)

(0.013) (0.017)

Center or preK 0.615+ 1.490**

0.143 0.732***

(0.311) (0.511)

(0.107) (0.150)

Assessment Covariates

Language (1 = Spanish) 27.512*** 34.850***

-1.908*** 23.928***

(0.774) (4.302)

(0.284) (0.916)

Timing 0.018*** 0.044***

0.024*** 0.018**

(0.004) (0.007)

(0.005) (0.006)

School Covariates

Sector (1 = public) -1.414*** -0.615

-0.088 -1.035**

(0.404) (0.685)

(0.236) (0.379)

Title1 (1 = yes)

-0.001 -0.209

(0.149) (0.240)

Size

0.099 0.034

(0.081) (0.119)

Minority representation

-0.004 -0.011**

(0.003) (0.004)

Constant -7.526** -25.185***

4.896*** 4.706**

(2.637) (3.576)

(1.017) (1.549)

Pseudo R2

0.357 0.323

0.592 0.612

Observations 6,320 6,320 16,460 16,460

Standard errors in parentheses

*** p<0.001, ** p<0.01, * p<0.05, + p<0.1

29

Table 1. Descriptive Statistics for Prekindergarten and Kindergarten Data

ECLS-B ECLS-K

Obs Mean Std. Dev. Min Max Obs Mean Std. Dev. Min Max

Dependent Variables

Math 6610 44.14 10.45 10.85 69.69

17310 27.72 8.87 7.44 59.34

Reading 6600 44.68 14.78 12.39 82.48

16580 32.14 10.05 11.00 70.80

Endogeneity Control

Previous math 6340 29.38 10.07 9.87 65.74

17120 19.61 7.38 6.65 59.82

Previous reading 6360 25.53 10.64 11.65 80.29

16120 22.26 7.85 10.08 60.00

Nonprofit Counts

Competitors 6670 10.64 9.27 0.00 65.00

17490 7.96 7.70 0.00 68.00

Supporters

Interveners 6670 38.78 51.05 0.00 1098.00

17490 35.10 61.11 0.00 1084.00

Youth developers 6670 5.42 5.18 0.00 76.00

17490 3.92 4.31 0.00 40.00

Community Covariates

Community SES 6680 0.00 0.79 -1.41 6.34

17500 0.00 0.85 -1.48 4.06

Urbanicity (1 = urban) 6560 0.83 0.38 0.00 1.00

17500 0.42 0.49 0.00 1.00

South 6680 0.36 0.48 0.00 1.00

17500 0.32 0.47 0.00 1.00

Children (6 and under) 6680 10510.16 8950.36 0.00 70895.00

17500 9939.85 8126.53 0.00 62892.00

Square miles 6670 764.17 1816.08 0.00 65362.15

17490 642.15 1347.73 0.00 22210.83

Child Covariates

Parental education 6670 5.01 2.05 1.00 9.00

17170 2.94 1.17 1.00 5.00

Low income 6680 0.44 0.50 0.00 1.00

17500 0.39 0.49 0.00 1.00

Center or preK 6640 0.96 1.00 0.00 4.00

16820 0.58 0.49 0.00 1.00

Age (months) 6680 68.63 4.74 57.20 86.00

17490 68.46 4.46 45.77 96.50

Gender (1 = female) 6680 0.50 0.50 0.00 1.00

17500 0.49 0.50 0.00 1.00 Table continued on next page.

30

Table 1. (Continued). Descriptive Statistics for Prekindergarten and Kindergarten Data

ECLS-B ECLS-K

Obs Mean Std. Dev. Min Max

Obs Mean Std. Dev. Min Max

Child Covariates

Race

White 6670 0.41 0.49 0.00 1.00 17470 0.57 0.50 0.00 1.00

African American 6670 0.16 0.36 0.00 1.00

17470 0.15 0.35 0.00 1.00

Hispanic 6670 0.20 0.40 0.00 1.00

17470 0.18 0.38 0.00 1.00

Other 6670 0.12 0.33 0.00 1.00

17470 0.11 0.32 0.00 1.00

Assessment Covariates

Language (1 = Spanish) 6651 0.00 0.02 0.00 1.00

17500 0.04 0.18 0.00 1.00

Timing (in days) 6680 79.84 49.49 15.00 530.00

17500 65.98 16.45 8.00 128.00

School Covariates

Sector (1 = public) 6450 0.88 0.33 0.00 1.00

17500 0.79 0.41 0.00 1.00

Title 1 receipt (1 = yes)

15030 0.60 0.49 0.00 1.00

Size

17330 3.29 1.16 1.00 5.00

Percent minority 17070 38.24 35.10 0.00 100.00

31

Table 2. Mean IRT Scores by Number of Nonprofits by Group

* p < 0.05, + p < 0.1 for differences between 0 and 1 to 5 and 0 and 5+.

Mean IRT Score (SD)

School Entry (ECLS-B) Kindergarten Year (ECLS-K)

Number of Nonprofits 0 1 to 5 5+ 0 1 to 5 5+

Math

Competitors 41.65 -10.14 43.30* -10.14 44.76* -10.56

27.39 -8.5 27.61 -8.73 27.64 -9.01

Supporters

Interveners 42.41 -9.54 42.69 -10.09 44.39* -10.51

27.35 -8.51 27.56 -8.73 27.63 -8.9

Youth developers 43.15 -9.98 43.97* -10.31 44.65* -10.76

27.4 -8.84 28.04* -8.89 26.90* -8.79

Reading

Competitors 42.52 -12.82 43.39 -14.15 45.45* -15.21

31.28 -9.85 31.66 -9.9 32.42* -10.16

Supporters

Interveners 41.7 -13.42 42.87 -13.33 45.00* -15.01

30.58 -9.79 31.23 -9.79 32.23* -10.09

Youth developers 43.53 -14.05 44.45+ -14.47 45.27* -15.41 31.35 -10.21 32.29* -9.98 31.99* -10.05

32

Table 3. Results of Models Predicting Math Scores by Nonprofit Count

*** p < 0.001, ** p < 0.01, * p < 0.05, + p < 0.1

Note: Models 2 and 3 contain controls for community SES, urbanicity, south, total children in zipcode area, total squaremiles in zipcode

area, parental education ,poverty status, center or prek care, gender of child, race of child, age of child, public/private school, assessment

timing, and language of assessment; as well as Title 1 receipt, school size, and percent school minority representation for ECLS-K models

β Coefficients (SE)

School Entry (ECLS-B)

Kindergarten Year (ECLS-K)

(1) (2) (1) (2) (3) (4)

Nonprofit Count

Competitors 0.256 -0.240

-0.107 -0.054 -0.240* -0.055

(0.281) (0.257)

(0.087) (0.088) (0.099) (0.087)

Supporters

Interveners -0.073 0.465*

-0.066 -0.066 -0.065 -0.067

(0.225) (0.211)

(0.076) (0.084) (0.082) (0.083)

Youth developers -0.312 -0.059

0.122 0.118 0.117 -0.009

(0.253) (0.225)

(0.091) (0.095) (0.094) (0.096)

Previous Math 0.260*** 0.212***

0.977*** 0.923*** 0.923*** 0.923***

(0.022) (0.019)

(0.010) (0.011) (0.011) (0.010)

Low Income (1 = yes)

-2.043***

-0.313* -0.981*** -0.590***

(0.383)

(0.122) (0.203) (0.149)

Family Income Interactions

Competitors x low income

0.428***

(0.127)

Youth developers x low income

0.299*

(0.133)

Constant 36.740*** -7.526**

9.023*** 4.896*** 5.220*** 5.020***

(0.823) (2.637)

(0.306) (1.017) (1.017) (1.025)

Pseudo R2

0.149 0.357 0.576 0.592 0.593 0.593

N 6,320 6,320

16,460 16,460 16,460 16,460

33

Table 4. Results of Models Predicting Reading Scores by Nonprofit Counts

*** p < 0.001, ** p < 0.01, * p < 0.05, + p < 0.1

Note: Models 2 and 3 contain controls for community SES, urbanicity, south, total children in zipcode area, total squaremiles

in zipcode area, parental education ,poverty status, center or prek care, gender of child, race of child, age of child,

public/private school, assessment timing, and language of assessment; as well as Title 1 receipt, school size, and percent school

minority representation for ECLS-K models

β Coefficients (SE)

School Entry (ECLS-B)

Kindergarten Year (ECLS-K)

(1) (2) (3)

(1) (2) (3)

Nonprofit Count

Competitors 0.023 -0.616+ -0.582

0.044 -0.055 -0.052

(0.383) (0.368) (0.367)

(0.136) (0.137) (0.137)

Supporters

Interveners 0.331 0.466 0.467

-0.034 -0.011 0.121

(0.346) (0.311) (0.307)

(0.134) (0.132) (0.144)

Youth developers -0.542 -0.051 0.353

0.105 0.343* 0.346*

(0.388) (0.332) (0.350)

(0.154) (0.145) (0.145)

Previous Reading 0.215*** 0.181*** 0.182***

0.757*** 0.563*** 0.564***

(0.029) (0.024) (0.024)

(0.018) (0.018) (0.018)

Low Income (1 = yes)

-2.798*** -1.541+

-1.793*** -0.919*

(0.583) (0.871)

(0.164) (0.380)

Family Income Interactions

Interveners x low income

-0.305*

(0.129)

Youth developers x low income

-0.958*

(0.433)

Constant 38.523*** -25.185*** -25.777***

14.934*** 4.706** 4.282**

(1.135) (3.576) (3.517)

(0.546) (1.549) (1.553)

Pseudo R2

0.102 0.323 0.324 0.531 0.612 0.612

N 6,320 6,320 6,320 16,460 16,460 16,460

34

Figure 1: Influences on Child Outcomes during the Transition to School

Child

Wellbeing

Families

Schools

Nonprofits

35

Figure 2. Predicted Kindergarten Math Test Scores in ECLS-K, by Competitor

Nonprofits and Family Income

26.8

27.0

27.2

27.4

27.6

27.8

28.0

28.2

28.4

Zero Three

Pre

dic

ted

Kin

der

gart

en M

ath

Tes

t Sc

ore

s

Competitor Nonprofits

Children from Low-Income Families Children from Families not Low-Income

36

Figure 3. Predicted Kindergarten Math Test Scores in ECLS-K, by Youth Developer

Nonprofits and Family Income

26.8

27.0

27.2

27.4

27.6

27.8

28.0

28.2

28.4

Zero Three

Pre

dic

ted

Kin

der

gart

en Y

ear

Mat

h S

core

Youth Developer Nonprofits

Children from Low-Income Families Children from Families not Low-Income

37

Figure 4. Predicted School Entry Reading Test Scores in ECLS-B, by Youth Developer

Nonprofits and Family Income

40.0

41.0

42.0

43.0

44.0

45.0

46.0

47.0

48.0

Zero Three

Pre

dic

ted

Sch

oo

l En

try

Rea

din

g Sc

ore

Youth Developer Nonprofits

Children from Low-Income Families Children from Families not Low-Income

38

Figure 5. Predicted Kindergarten Reading Test Scores by Intervener Nonprofits and

Family Income

30.0

30.5

31.0

31.5

32.0

32.5

33.0

33.5

Zero Three

Pre

dic

ted

Kin

der

gart

en R

ead

ing

Test

Sco

res

Intervener Nonprofits

Children from Low-Income Families Children from Families not Low-Income

39

References

Alexander, Karl L., Dorris R. Entwisle and Linda S. Olson. 2007. “Lasting consequences

of the summer learning gap,” American Sociological Review. 72(2): 167-180.

Allard, Scott W. 2009. Out of Reach: Place, Poverty, and the New American Welfare

State. New Haven, CT: Yale University

Altonji, Joseph G., and Richard K. Mansfield. 2011. "The Role of Family, School, and

Community Characteristics in Inequality in Education and Labor-Market

Outcomes." Pp. 339-58 in Whither Opportunity? Rising Inequality, Schools, and

Children's Life Chances edited by Greg J. Duncan and Richard Murnane. New

York: Russell Sage Foundation.

Anheier, Helmut K. 1954. Nonprofit Organizations: An Introduction: Theory,

Management, Policy. New York, NY: Routledge.

Armbruster, Bonnie B., Fran Lehr, and Jean Osborn. 2001. Put reading first: the research

building blocks for teaching children to read : kindergarten through grade 3.

Washington, D.C.: National Institute for Literacy, National Institute of Child

Health and Human Development, U.S. Dept. of Education

Arum, Richard. 2000. "Schools and Communities: Ecological and Institutional

Dimensions," Annual Review of Sociology. 26:395-418.

Barnett, Steven W. and C. R. Belfield. 2006. “Early Childhood Development and Social

Mobility.” The Future of Children, 16(2): 73-98

40

Bates, Littisha A. and Jennifer Glick. 2013. “Does it Matter if Teachers and Schools

Match the Student? Racial and Ethnic Disparities in Problem Behaviors.” Social

Science Research, 42(5), 1180-1190.

Bertalanffy, Ludwig von. 1969. General system theory: foundations, development,

applications. New York, NY: G. Braziller.

Bettinger, Eric P. 2012. “Paying to Learn: The Effect of Financial Incentives on

Elementary School Test Scores.” The Review of Economics and Statistics, 94(3):

686-698

Bielefeld, Wolfgang. 2000. “Metropolitan Nonprofit Sectors: Findings from NCCS

Data.” Nonprofit and Voluntary Sector Quarterly, 29(2): 297-314

Ceci, Stephen J. and Paul B. Papierno. 2005. “When the “Have-Nots” Gain but the

“Haves” Gain Even More.” American Psychologist,60(2): 149-160.

Crosnoe, Robert and Aprile Benner. (in press). “Children at School.” In Handbook of

Child Psychology and Developmental Science, Vol. 4: Ecological Settings and

Processes, edited by Marc Bornstein & Tama Leventhal (series editor: Richard

M. Lerner). New York, NY: Wiley

Crosnoe, Robert, Claude Bonazzo, and Nina Wu. (in press). Healthy Learners: Poverty,

Immigration, and Opportunity in Early Childhood Education. New York, NY:

Teachers College Press.

Darensbourg, Alicia M. and Jamilia J. Blake. 2014. “Examining the Academic

Achievement of Black Adolescents: Importance of Peer and Parental Influences.”

Journal of Black Psychology, 40(2), 191-212.

41

Dee, Thomas and Brian Jacob. 2010. “Evaluating NCLB: Accountability has Produced

Substantial Gains in Math Skills but not in Reading.” Education Next, 10(3): 54

Duncan, Greg J and Richard J Murnane. 2014. “Growing Income Inequality Threating

American Education: Rising Economic and Social Inequality has Weakend

Neighborhoods and Families in Ways that Make Effective School Reform More

Difficult” Phi Delta Kappan. 95(6): 8.

Duncan, Greg J. and Katherine Magnuson. 2013. “Investing in Preschool Programs.” The

Journal of Economic Perspectives, 27(2): 109-132.

Fuller, Bruce. 2007. Standardized Childhood: The Political and Cultural Struggle over

Early Education. Stanford, CA: Stanford University Press.

Garrett, Thomas and Russell Rhine. 2010. “Government Growth and Private

Contributions to Charity.” Public Choice, 143(1/2): 103-120.

Glazerman, Steven, Dan Levy, and David. Myers. 2000. “Nonexperimental versus

experimental estimates of earnings impacts,” Annals of the American Academy of

Political and Social Science, 589: 63-85.

Heckman, James J. 2006. “Skill Formation and the Economics of Investing in

Disadvantaged Children.” Science, 312(5782):1900-1902.

Heckman, James J. 2000. “Policies to Foster Human Capital.” Research in Economics,

54(1): 3-56.

Invernizzi, Marcia, Catherine Rosemary, Connie Juel and Herbert C. Richards. 1997.

“At-Risk Readers and Community Volunteers: A 3-Year Perspective.” Scientific

Studies of Reading, 1(3): 277-300.

42

Katz, Hagai. 2014. “Poverty and Other Factors Affecting the Location of Nonprofit

Human Service Organizations.” International Journal of Social Work and Human

Services Practice Horizon Research Publishing, 2(5): 163-172

Leana, Carrie R. and Frits K. Pil. 2006. ”Social Capital and Organizational Performance:

Evidence from Urban Public Schools.” Organization Science, 17(3): 353-366

Lee, Valerie E., and David T. Burkam. 2002. Inequality at the starting gate: Social

background differences in achievement as children begin school. Washington,

DC: Economic Policy Institute.

Mason, Jana M. 1980. “When Do Children Begin to Read: An Exploration of Four Year

Old Children's Letter and Word Reading Competencies” Reading Research

Quarterly, 15(2): 203-227.

McDouble, Lindsey M. and Marcas Lam. 2014. “Individual- and Community-Level

Determinants of Public Attitudes Towards Nonprofit Organizations.” Nonprofit

and Voluntary Sector Quarterly, 43(4): 672-692.

McKown, C. and Weinstein, R. S. 2002. “Modeling the Role of Child Ethnicity and

Gender in Children's Differential Response to Teacher Expectations.” Journal of

Applied Social Psychology, 32: 159–184.

National Center for Charitable Statistics and Independent Sector. 2002. The New

Nonprofit Almanac and Desk Reference. Waldorf, MD: Jossey-Bass.

National Science Board. 2014. Science and Engineering Indicators 2014. Arlington VA:

National Science Foundation (NSB 14-01).

43

OECD. 2007. Education at a Glance 2007: OECD Indicators.

http://www.oecd.org/document/30/0,3343,en 2649 39263238 39251550 1 1 1

1,00.html.

Pianta, Robert, and Daniel Walsh. 1996. High-risk children in schools: Constructing

sustaining relationships. New York, NY: Routledge.

Pianta, Robert C., Martha J. Cox,. and Kyle L. Snow. 2007. School Readiness and the

Transition to Kindergarten in the Era of Accountability. Baltimore, MD: Paul H.

Brooks

Reich, Rob. 2005. “A Failure of Philanthropy: American charity shortchanges the poor,

and public policy is partly to blame.” Stanford Social Innovation Review, 3(4):

24-33.

Ritter, Gary W., Joshua H. Barnett, George S. Denny and Ginger R. Albin. 2009. “The

Effectiveness of Volunteer Tutoring Programs for Elementary and Middle School

Students: A Meta-Analysis.” Review of Educational Research, 79(7): 3-38

Rose-Ackerman, Susan. 1996. “Altruism, Nonprofits, and Economic Theory.” Journal of

Economic Literature, 34(2): 701-728.

Salamon, Lester M., Leslie C. Hems, and Kathryn Chinnock. 2000. “The Nonprofit

Sector: For What and for Whom?” Working Papers of the Johns Hopkins

Comparative Nonprofit Sector Project, 37. Baltimore: The Johns Hopkins Center

for Civil Society Studies.

44

Sénéchal, Monique and Laura Young. 2008. “The Effect of Family Literacy Interventions

on Children's Acquisition of Reading from Kindergarten to Grade 3: A Meta-

Analytic Review” Review of Educational Research, 78(4): 880-907

Shonkoff, Jack P. and Deborah Phillips. 2000. From Neurons to Neighborhoods: The

Science of Early Childhood Development. Washington, DC: National Academy

Press.

Snow, Catherine E., Susan M. Burns, and Peg Griffin. 1998. Preventing Reading

Difficulties in Young Children. Washington, D.C.: National Academy Press.

U.S. Department of Education. 2002. No Child Left Behind (NCLB) Act of 2001, Pub. L.

No. 107-110, 115, Stat. 1425 .

Varnhagen, Connie K., Frederick J. Morrison, and Robin Everall. 1994. “Age and

schooling effects in story recall and story production.” Developmental

Psychology, 30, 969-979.

45

Vita

Robert (b. 1988, Philadelphia) focuses his work on the sociology of education with an

interest in public-school/nonprofit partnerships, stratification, and inequality.

Permanent email: rwress@utexas.edu This thesis was typed by Robert Wayne Ressler